Reproducibility is essential in informatics and AI research. Every research article in JMIDH must include both a Data Availability Statement and a Code Availability Statement. Editors check these statements before acceptance.

Code

•     Code used to process data, train and evaluate models, and generate the main results should be deposited in a public repository (for example GitHub or GitLab) and archived with a persistent identifier (for example through Zenodo), under an open-source licence such as MIT, Apache 2.0 or GPL. The archived version and its DOI must be cited in the article.

•     Software and Tool papers must make their code publicly available with documentation and installation instructions; this is a condition of publication.

•     Where code cannot be shared (for example because it is proprietary), authors must explain why and provide enough detail — algorithms, pseudo-code, hyperparameters and software versions — for others to reproduce the work. Studies of commercial products must describe the product version evaluated.

•     For review, code must be provided in anonymised form (see Author Instructions).

Models

Authors are encouraged to share trained model weights or provide an accessible way to apply the model (for example an online calculator or API), together with a model card. Where the full model equation is simple (for example a logistic regression), it must be given in full so that others can validate it. If model sharing is restricted because of privacy, security or commercial reasons, this must be explained.

Data

•     De-identified data should be shared in a recognised repository (for example Zenodo, Figshare, Dryad, PhysioNet, Harvard Dataverse or OSF) when consent and ethics approval allow.

•     When data cannot be shared publicly — as is often the case for patient records — authors must explain the restriction and how qualified researchers can request access (for example through the data custodian or a data access committee).

•     Data from credentialed public databases such as MIMIC may not be redistributed; authors should cite the dataset and share code that allows others with access to reproduce the analysis.

•     Synthetic data may be shared as a substitute for real data, but its generation method and limitations must be described, and it must not be presented as real data.

•     Data from clinical trials must be accompanied by a data sharing statement in line with ICMJE.

Statements

Example of a Data Availability Statement: ‘The de-identified dataset is available at [repository, DOI] under a CC BY 4.0 licence.’ or ‘Patient data cannot be shared publicly due to [reason]; requests can be made to [custodian] subject to [conditions].’ Example of a Code Availability Statement: ‘All code is available at [repository URL] and archived at [Zenodo DOI] under the MIT licence.’

Editors may request code, models or data during review or after publication to examine questions about the results. Failure to provide them when concerns are raised may lead to rejection, an expression of concern or retraction.